Design and implement specialized Claude Code subagents and multi-phase workflow orchestrators.
The Write Agent skill provides a comprehensive framework for creating, refactoring, and managing custom AI agents within the Claude Code environment. It helps developers navigate the architectural decision between single-purpose subagents for tasks like code review and complex workflow orchestrators for multi-step feature development. By enforcing principles like token efficiency, clear output contracts, and robust delegation triggers, this skill ensures that custom agents remain reliable, focused, and maintainable as repository complexity grows.
主な機能
011 GitHub stars
02Comprehensive procedures for refactoring existing agents without routing drift
03Guidance on tool permissions, task-tool orchestration, and model selection
04Architecture selection between subagents, orchestrators, and repo rules
05Standardized templates for reviewers, workers, and workflow coordinators
06Framework for defining robust input/output contracts for multi-agent systems
ユースケース
01Refactoring verbose legacy agent prompts to optimize token budget and attention
02Building a multi-phase feature development orchestrator with iteration gates
03Creating a specialized security reviewer subagent with read-only tool access
